mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Qian Zhang <zhq527...@gmail.com>
Subject Re: Review Request 65570: Attached/detached volume directory for task which has volume specified.
Date Tue, 13 Feb 2018 12:35:26 GMT


> On Feb. 13, 2018, 9:38 a.m., Vinod Kone wrote:
> > Can you add tests for this in a subsequent review please?

Yes, here it is: https://reviews.apache.org/r/65630/


> On Feb. 13, 2018, 9:38 a.m., Vinod Kone wrote:
> > src/slave/slave.cpp
> > Lines 1109-1111 (patched)
> > <https://reviews.apache.org/r/65570/diff/2/?file=1955934#file1955934line1109>
> >
> >     I don't think we need this check. We want to show non PV directories too in
the UI.

That was actually my original solution, but after more consideration and testing, I found
we still need this check. The reason is, if we do not have this check, that means we will
do attach as long as the task's ContainerInfo has a `SANDBOX_PATH` volume with type `PARENT`,
but such attach will fail (see below) because the executor's volume directory is not created
yet at the moment when `attachTaskVolumeDirectory` is called, it will actually be created
by the `volume/sandbox_path` isolator when launching the nested container.
```
E0212 22:23:08.905462 10350 slave.cpp:1009] Failed to attach '/home/qzhang/opt/mesos/slaves/54de35d5-250f-4bf9-9444-3633c5e66e7d-S0/frameworks/54de35d5-250f-4bf9-9444-3633c5e66e7d-0000/executors/default-executor/runs/0ea4dc53-e543-42c4-8bb5-c5b667eee4c0/foo'
to virtual path '/home/qzhang/opt/mesos/slaves/54de35d5-250f-4bf9-9444-3633c5e66e7d-S0/frameworks/54de35d5-250f-4bf9-9444-3633c5e66e7d-0000/executors/default-executor/runs/0ea4dc53-e543-42c4-8bb5-c5b667eee4c0/tasks/test/xxx':
Failed to get realpath of '/home/qzhang/opt/mesos/slaves/54de35d5-250f-4bf9-9444-3633c5e66e7d-S0/frameworks/54de35d5-250f-4bf9-9444-3633c5e66e7d-0000/executors/default-executor/runs/0ea4dc53-e543-42c4-8bb5-c5b667eee4c0/foo':
No such file or directory
```
So I think we still need this check which will ensure the executor's volume directory exists
before we do the attach, the executor's volume directory will actually be created by the `filesystem/linux`
isolator when launching the executor.


- Qian


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/65570/#review197359
-----------------------------------------------------------


On Feb. 13, 2018, 5:20 p.m., Qian Zhang wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/65570/
> -----------------------------------------------------------
> 
> (Updated Feb. 13, 2018, 5:20 p.m.)
> 
> 
> Review request for mesos, Gilbert Song and Vinod Kone.
> 
> 
> Bugs: MESOS-8565
>     https://issues.apache.org/jira/browse/MESOS-8565
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Attached/detached volume directory for task which has volume specified.
> 
> 
> Diffs
> -----
> 
>   src/slave/slave.hpp 30151c4886e12e9183a971b86b854e28a8ca1b39 
>   src/slave/slave.cpp f98f37321872d090176b7cc50873fc3c627773f5 
> 
> 
> Diff: https://reviews.apache.org/r/65570/diff/3/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Qian Zhang
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message